Rapid defluoridation of drinking water by calcium carbonate nanoadsorbent: characterization, adsorption studies and application to real samples’ treatment

Fluoride (F⁻), an element essential in low concentrations for healthy teeth and bone development, can lead to adverse health issues when its intake exceeds permissible (1.5 mg L⁻¹, World Health Organization (WHO); ≤1.5 mg L⁻¹, National Standards for Drinking Water Quality, Pakistan) or in some cases desirable limits (1.0 mg L⁻¹, WHO) (Mohapatra et al. 2009). Health issues associated with high fluoride levels include dental caries; dental fluorosis; skeletal, brain, and kidney damage; dietary allergies; changes in deoxyribonucleic acid (DNA) structure; and stomach problems (Reardon & Wang 2000; Wang et al. 2004; Fawell et al. 2006; Oruc 2008; Barbier et al. 2010; Kut et al. 2016). Adverse impacts of fluoride contamination have also been observed in aquatic organisms (Camargo 2003).

According to an estimate, fluoride levels have considerably increased in the groundwaters of Central Asian countries including Pakistan (Bashir et al. 2013) leading to reported health concerns like fluorosis, impaired mental as well as physical development, bone deformities, hip fractures and kidney problems (Azizullah et al. 2011). Elevated levels of F⁻ in water can be attributed to anthropogenic sources (Ullah et al. 2009) and slow dissolution of fluoride-containing rocks in groundwater (Diesendorf et al. 1998) along with effluents from different industries (Banks et al. 1995; Edmunds & Smedley 2013). Exposure to high fluoride levels from daily intake of fluorinated drinking water (DW) puts people at high risk of encountering health issues. Worldwide fluoride consumption has increased enormously either from intake of fluoride-supplemented water or use of natural fluoridated water. There are over 30 countries consuming fluorinated water at levels that exceed the WHO permissible limit (Tiemann 2011, 2013). Hence, the United States Environment Protection Agency (EPA) has classified fluoride as a major water contaminant across the world (EPA 2009).

Different treatment technologies are currently available for defluoridation of water, such as adsorption (Bhatnagar et al. 2011; Bia et al. 2012), reverse osmosis, ion-exchange (Mohapatra et al. 2009), conventional (Gogoi et al. 2015), and electro coagulation–flocculation (Holt et al. 2005; Zuo et al. 2008; Emamjomeh & Sivakumar 2009) but these techniques are associated with one or more disadvantages such as length of time required, high maintenance cost, and production of environmental toxic by-products etc. (Maheshwari 2006; Ayoob et al. 2008; Li et al. 2011; Goswami & Purkait 2012; Jagtap et al. 2012; Rafique et al. 2012).

Recent times have seen the upsurge of nanotechnology application in many areas. Nanotechnology has proved a reliable, cost-effective, and highly efficient technique to treat DW (Rao et al. 2009; Ali 2012; Pontie et al. 2013; Kumar & Tomar 2014). For defluoridation of DW, adsorption is considered to be an efficient option compared to other technologies owing to simplicity of steps, operation, and maintenance cost (Bhatnagar et al. 2011; Jagtap et al. 2012). Moreover, nanoadsorbents result in high removal efficacy because of improved properties (Chang et al. 2006; Mansoori et al. 2008; Rao et al. 2009; Devi et al. 2014; Kumar & Tomar 2014; Adak et al. 2017b). Different nanoadsorbents, simple to complex, used for treating groundwater and DW with varying adsorption capacities include iron (Fe)–titanium (Ti) bimetallic oxide/magnetite (Fe₃O₄) (Zhang et al. 2014), aluminum oxide (Al₂O₃) (Tangsir et al. 2016; Hafshejani et al. 2017), magnesium oxide (MgO) (Oladoja et al. 2015), Al(III)–Fe(III)–La(III) trimetallic oxide (Adak et al. 2017b), iron oxide (Zhang et al. 2017), Fe–Ag magnetic binary oxide (Azari et al. 2015) and FeMgLa trimetal nanocomposite (Chen et al. 2018) and others (Adak et al. 2017a; Sani et al. 2017;Yan et al. 2017; Chen et al. 2018).

The present study was designed to explore new low cost nanoadsorbents for effective and efficient defluoridation of DW in short time without requirements for stringent conditions either in synthesis or in adsorption process.

Calcium chloride (CaCl₂·2H₂O) and sodium fluoride (NaF) were obtained from Fluka while sodium carbonate (Na₂CO₃) was procured from Sigma-Aldrich. Concentrated hydrochloric acid (HCl) was obtained from the Pakistan Council of Scientific and Industrial Research, Laboratories Complex. Distilled water was used for preparation of standard and experimental solutions.

For the preparation of calcium carbonate (CaCO₃) nanoadsorbent a simple modified precipitation method was adopted (Ghadam & Idrees 2013). To 0.15 M calcium chloride (16.6 g), taken in a two-necked round-bottom flask, 0.4 M NaCO₃ (42.4 g) solution was added drop wise, with continuous stirring over a period of 24 h. White precipitates of CaCO₃ formed were filtered, washed with distilled water followed by oven drying at 105 °C for 3–4 h. The precipitates were then annealed in cube furnace at 650 °C for 4–5 h to obtain off-white powder.

Nano-adsorbency was characterized using Fourier transform infrared spectroscopy (FT-IR), powder X-ray diffraction (XRD), scanning electron microscopy (SEM)/energy-dispersive X-ray spectroscopy (EDX) and atomic force microscopy (AFM) in order to determine phase, particle size, purity, and composition. Potassium bromide (KBr) pellets of sample were used for FT-IR spectrum (Bruker FT-IR). SEM/EDX (S3700N, Hitachi, Japan) and AFM (AFM 5500 Agilent, USA) images were taken for morphology and particle size. Powder XRD (PANalytical) was carried out using continuous mode.

Stock solution (1,000 mg L⁻¹) of the fluoride ion (F⁻) was prepared from NaF in distilled water, and further dilutions were made according to the requirements.

Batch adsorption studies were carried out to determine the effectiveness of adsorbent for uptake of F⁻ from aqueous solution. The experimental parameters were optimized for maximum adsorption by varying contact time, temperature, initial pH, initial F⁻ concentration, and adsorbent dose.

Kinetic, adsorption, and diffusion models were applied to correlate the data, in line with earlier researches carried out to determine adsorption efficacy of new adsorbents (Jain et al. 2010; Gusmão et al. 2013; Keränen et al. 2015).

Both models are used to assess whether the retention process is controlled by pore (Equation (7)) or particle diffusion (Equation (8)).

A total of 30 DW samples were collected from 30 locations in the Lahore study area. Water supply in these areas is through tube wells installed in all parts of the district and is pumped directly into the distribution systems. Composite sampling technique was applied for sample collection from water taps by taking five random tap water samples (1 L) over a period of 8 h and then combining to get a composite sample. The samples were collected, coded (Table 1), preserved in bottles, and stored at 4 °C prior to analysis. The samples were analyzed for F⁻ using ion electrode prior to treatment; the results of analysis are presented in Figure 1.

DW samples (100 mL) were taken in a conical flask and, after maintaining pH between 7 and 8 (using 0.1 N HCl or 0.1 N NaOH), synthesized CaCO₃ nanoadsorbent (0.1 g) was added to the flask. The contents were stirred for 10 min at room temperature (RT ∼ 30 °C) followed by Cannula filtration. The treated samples (filtrates) were analyzed for F⁻ ion concentration using ion selective electrode (930, Spectrum Scientific).

Prepared nanoparticles were then characterized by the following techniques:

The initial conformation for the synthesis of CaCO₃ was performed using FT-IR, the spectrum of which is presented in Figure 2(a). The spectrum shows two peaks, with the strongest peak appearing at 1,405.78 cm⁻¹ and medium intensity peak at 874.69 cm⁻¹. Both peaks are attributed to bending and stretching vibrations of the O–C–O bond (Abdolmohammadi et al. 2012). The spectra obtained is in good agreement with the calcite characteristics peaks (Abdolmohammadi et al. 2012; Kirboga & Oner 2013).

The powder XRD of CaCO₃ was performed for confirmation of phase. The diffractogram showed presence of peaks indicative of calcite (Kontoyannis & Vagenas 2000), with few peaks fitting to Ca(OH)₂ (Taglieri et al. 2013) as impurity, marked by asterisk in Figure 2(b). The diffractogram is characterized by well-defined peaks, with broadening indicating the formation of nano-sized material (Uvarov & Popov 2007).

The morphology of prepared nanoparticles, examined through SEM images (Figure 3(a)), exhibited spongy appearance of synthesized material, which might result in high adsorption capacity of F⁻. Elemental composition (weight %) as determined from EDX (Figure 3(b)) was 35.83% Ca, 14.35% C, and 49.82% O, corresponding to ratio of 1:1.3:3.5 (Ca:C:O), which is in good agreement to original ratio of 1:1:3, hence confirming formation of CaCO₃.

The nano-size of CaCO₃ particles prepared was confirmed using AFM. Figure 3(c) and 3(d) presents the AFM images in both 2D and 3D view, showing particle size of 14.6 nm. The image shows narrow distribution of the particles, which have a distorted oblong shape.

Adsorption potential of nano-CaCO₃ was evaluated by optimizing parameters like contact time, temperature, initial pH, initial F⁻ concentration, and adsorbent dose.

Room temperature batch adsorption studies were done as a function of time. The plot (Figure 4) of adsorption capacity (mg g⁻¹) against time (2–115 min) indicates rapid adsorption of F⁻ onto the nanoadsorbent, resulting in attainment of equilibrium within 10 min. The optimized time is much less than that already reported for other nanoadsorbents (Sundaram et al. 2008, 2009; Chen et al. 2012; Hafshejani et al. 2017).

Fitting of experimental data applied to kinetic models (general (Equation (3)), pseudo-first order (Equation (4)), and pseudo-second order (Equation (5))) and related parameters calculated in each case are given in Figure 4(a) and Table 3, respectively. The models' fitting results showed the best-fit model to be the general order compared to the rest of models based on high R² value, smaller standard deviation, and reduced-χ² value (Equation (6)) (Saucier et al. 2015).

The plots for particle (Equation (8)) and intra-particle (Equation (7)) diffusion models are given in Figure 4(b) and 4(c) and resulting parameters are shown in Table 2. Diffusion models give information regarding the rate-limiting step and number of processes involved in adsorption (Ciopec et al. 2014). The plot q_t versus t^0.5 (Figure 4(c)) showed non-linear dependency over the total range, with the line not passing through the origin, indicating the main rate-limiting step to be controlled by boundary layer diffusion. The plot is fitted to three straight lines, suggesting occurrence of different sorption phenomena (Ciopec et al. 2014), and depicts rapid initial adsorption owing to fast uptake of ions onto CaCO₃ surface, which is saturated later and leads to slower adsorption rates once equilibrium is approached. The third step corresponds to the slow process of transportation of F⁻¹ into adsorbent pores (Figure 4(c)) (Ciopec et al. 2014). Hence the phenomenon is controlled predominantly by particle rather than pore diffusion, as also confirmed by high R² value (Table 3).

Temperature plays an important part in adsorption process considering that endothermic processes require more energy input, which makes water purification process costly. For current studies, the adsorption studies were performed by altering temperature from 30 °C to 100 °C while keeping other parameters constant (Ci = 500 mg L⁻¹). Figure 5(a) shows a very small incremental decrease in adsorption capacity with an increase in temperature making the adsorption phenomenon temperature-independent. Further studies were therefore performed at room temperature. This temperature-independent adsorption enhances effectiveness of the process as no extra energy is required.

To determine impact of pH on adsorption capacity of CaCO₃, room temperature experiments were conducted in pH range of 1–14 with 500 mg L⁻¹ fluoride concentration using 0.1 g adsorbent. High F⁻ adsorption was observed at much broader range, i.e. full studied range (Figure 5(a)). Hence, pH maintenance is not required prior to water treatment. In contrast, earlier studies carried out using nanoadsorbents showed maximum adsorption either at low pH or in a narrow range (Li et al. 2001; Reyes Bahena et al. 2002; Tembhurkar & Dongre 2006; Sundaram et al. 2008; Bhatnagar et al. 2011).

The effect of nanoadsorbent dose was investigated by varying its amount from 0.01 to 0.125 g. Results (Figure 5(a)) indicate that with an increase in adsorbent dose the percentage of fluoride adsorption rises and reaches equilibrium at 0.025 g. An increase in amount of adsorbent is directly proportional to surface area, leading to enhancement in adsorption percentage (Tembhurkar & Dongre 2006). For further studies this optimized dose was used.

To determine the effect of initial adsorbate concentration on adsorption efficiency of nanoadsorbent, different F⁻ concentrations were treated with optimized adsorbent dose, i.e. 0.025 g, and the q_e value obtained thereafter showed optimized value at 800 mg L⁻¹. To get insight into phenomenon involved in this adsorption, various models were applied; plots and results of these studies are presented in Figure 5(b) and Table 3 respectively. The best-fit model was assessed from a high R² value, smaller SD, and reduced-χ² value.

The four adsorption models employed (Langmuir (Equation (9)), Freundlich (Equation (10)), Liu (Equation (11)), and R–P (Equation (12))) showed R_adj² values in range of 0.9537–0.9927 (Figure 5(b), Table 3). The data fits well to the Liu model, with the highest R_adj² and smallest reduced-χ² value and a q_m of 754.36 mg g⁻¹ (Table 3), which is very high compared to previously reported nanoadsorbents, i.e. 2.22, 2.84, 14.9, 47.0, and 91.04 mg g⁻¹ for iron–aluminum–cerium (Chen et al. 2009), nano-hydroxyapatite/chitin (Sundaram et al. 2009), Al₂O₃/CNT (Li et al. 2001), Fe–Ti oxide (Chen et al. 2012), magnetic core-shell Ce-Ti@Fe₃O₄ (Abo Markeb et al. 2017), respectively. The higher bonding of adsorbate with adsorbent is also denoted by high value of n_F obtained from the Freundlich isotherm (Figure 5(b), Table 3) (Khalid et al. 2015). The g value (close to unity) obtained in the R–P model implies monolayer adsorption behavior and compatibility with Langmuir rather than Freundlich as also indicated by the R_adj² value (El-Sikaily et al. 2007; Amrhar et al. 2015).

To assess the potential of nanoadsorbent in removing F⁻ from real water samples, 30 water samples were collected and treated under optimized conditions (contact time 15 min, room temperature, pH 7–8). The concentration of F⁻ was measured before and after treatment.

According to EPA (2009) the maximum contaminant level allowed is 4 mg L⁻¹, while according to the WHO water quality guidelines, the acceptable concentration is 1.5 mg L⁻¹. Out of 30 samples, only six samples exceeded the WHO limits and one sample was above the EPA limits. Only nine samples, with F⁻ concentration above 1.0 mg L⁻¹, were subjected to adsorption. The results (Figure 6) showed efficacy of nanoadsorbant by appreciable reduction in F⁻ level. The fluoride concentration in samples (6.497–0.999 mg L⁻¹) reduced to 0.057–0 mg L⁻¹ after treatment.

CaCO₃ nanoparticles (14.6 nm) were prepared by a simple co-precipitation route for the defluoridation of DW. Adsorption studies at room temperature showed high adsorption capacity, i.e. 725.21 mg g⁻¹, attained in a short time. Application for treatment of actual water samples also demonstrated 98–100% removal.

The current research dealt with exploiting the potential of prepared CaCO₃ nanoparticles toward F⁻ removal in simulated as well as real samples. The findings have opened up new dimensions for an easy and effective water defluoridation technique. In future, the studies can be extended to different dimensions to enhance its fruitfulness, as detailed below:

• Evaluating the effectiveness of prepared nanoparticles using continuous flow studies by varying the volume of water influx.

• Binding the CaCO₃ nanoparticles into some high surface area material for making an adsorbent bed in order to avoid the filtration challenge posed by use of nanoparticles.